Nindynar Rikatsih, Ahmad Afif Supianto
The existing classification methods are mostly designed to solve problems with multivariate time series data types although in general there is also univariate time series data attributes that are often encountered in the application of classification problems. This research raises the problem with univariate time series data type in posture reconstruction data classification. We propose a classification technique such as k-nearest neighbor for classification of univariate time series data type. We first collect the data of posture reconstruction. from k-NN method we determine that the number of k is 5 and doing the similarity measurement by using Euclidean Distance. We rank the results based on the number of k and the similarity measurement to get the final result. Based on the experiment of this research, the accuracy that we get from k-NN is 0.995 with the balance precision and recall. Therefore, classification method particularly k-NN can be used to solve the univariate time series problem in posture reconstruction. © 2018 IEEE.
Faculty of Computer Science, Brawijaya University, Malang, Indonesia
